Folding of a donor–acceptor polyrotaxane by using noncovalent bonding interactions

Abstract

Mechanically interlocked compounds, such as bistable catenanes and bistable rotaxanes, have been used to bring about actuation in nanoelectromechanical systems (NEMS) and molecular electronic devices (MEDs). The elaboration of the structural features of such rotaxanes into macromolecular materials might allow the utilization of molecular motion to impact their bulk properties. We report here the synthesis and characterization of polymers that contain π electron-donating 1,5-dioxynaphthalene (DNP) units encircled by cyclobis(paraquat-p-phenylene) (CBPQT⁴⁺), a π electron-accepting tetracationic cyclophane, synthesized by using the copper(I)-catalyzed azide-alkyne cycloaddition (CuAAC). The polyrotaxanes adopt a well defined “folded” secondary structure by virtue of the judicious design of two DNP-containing monomers with different binding affinities for CBPQT⁴⁺. This efficient approach to the preparation of polyrotaxanes, taken alongside the initial investigations of their chemical properties, sets the stage for the preparation of a previously undescribed class of macromolecular architectures.

The design of polymeric materials incorporating broad-scale, electromechanical motion such as those associated with electrochemically switchable, bistable [2]rotaxanes (1, 2) and bistable [2]catenanes (3, 4) might allow these nanoscopic phenomena to be harnessed and to impact macroscopic properties. New classes of processable materials, potentially capable of bulk actuation (5), the stimulus-driven release of therapeutics (6), or stimuli-responsive changes in wettability or electronic properties (7) will result. Linear polyrotaxanes are one such architecture of interest, with relatively few reported examples except for those using cyclodextrin (8) or cucurbituril (9) derivatives as the macrocyclic component. Polyrotaxanes based on donor–acceptor binding interactions, and featuring cyclobis(paraquat-p-phenylene) (CBPQT⁴⁺) as the π-accepting ring component, are highly desirable targets because its redox-controlled electromechanical motions have already been used in a range of nanoelectromechanical systems (NEMS) (2) and molecular electronic devices (MEDs) (10–12).

Such donor–acceptor polyrotaxanes have, thus far, been synthetically elusive, mostly because the template-directed “clipping” methodology used traditionally for the installation of CBPQT⁴⁺ rings is not efficient enough to obtain a high coverage of macrocycles encircling the polymer backbone. We recently developed, however, a “threading-followed-by-stoppering” protocol for the synthesis of donor–acceptor rotaxanes (13) and catenanes (14), which takes advantage of the high efficiency and excellent functional group tolerance of the Cu(I)-catalyzed azide–alkyne cycloaddition (CuAAC) (15, 16). Inspired by the growing use of this transformation for polymer synthesis and functionalization (17, 18), we have adapted this “click chemistry” approach to the synthesis of high-molecular-mass polymers containing π electron-donating 1,5-dioxynaphthalene (DNP) units along the polymer chain. Thereafter, CBPQT⁴⁺ rings were threaded onto the chains and bulky stoppers were attached, employing the CuAAC reaction once again.

One of the first unique properties observed in these polyrotaxanes is that they fold into a compact structure as a consequence of secondary intramolecular interactions. By virtue of the alternating nature of the A–A + B–B polymerization, in concert with the judicious design of the two DNP-containing monomers with differing binding affinities for CBPQT⁴⁺, a majority of the rings encircle the more favorable DNP recognition site, whereas the secondary DNP recognition sites stack on the outside of the rings. In this manner, the polymer adopts a folded conformation (Fig. 1) in a fashion reminiscent of the pleated secondary structure adopted by polymers of alternating π electron donors and acceptors in aqueous solution (19). This folding behavior has also been observed in more rigid polymers and oligomers of certain polyamides (20), as well as in m-phenylene ethylene (21) “foldamers” (22–25).

Fig. 1.

Schematic representation of the formation of polymer thread and the polyrotaxane from two DNP-containing monomers.

Synthetic Investigations

The two monomers used for the synthesis of the polymer thread are each derivatives of 1,5-bis(2-(2-hydroxyethoxy)ethoxy)naphthalene (BHEEN). Their differing binding affinities toward CBPQT⁴⁺ are a direct consequence of their terminal substitution: the propargyl ether-terminated glycol chains of 1 contain an extra oxygen atom relative to the diazide-terminated (26) diethylene glycol chains of 2. These oxygen atoms are key for the [C–H···O] interactions that strengthen the CBPQT⁴⁺–DNP complex (27, 28). The CuAAC polymerization was carried out (Scheme 1) by using a mixture of the two monomers in the presence of 0.05 equivalents of CuSO₄·5H₂O and 0.1 equivalent of ascorbic acid in DMF at room temperature. The number-average degree of polymerization X_n was controlled by employing a slight stoichiometric imbalance in the feed ratio of the two monomers, r, as described by Eq. 1 (29).

Scheme 1.

Reaction scheme illustrating the synthesis of the polyrotaxane 5·4nPF₆.

We varied the ratio of the two DNP-containing monomers to obtain (Table 1) polymers with different molecular mass—namely, 32 kDa, 53 kDa, and 181 kDa, respectively. In the case of polymers 3a and 3b, the diazide monomer 2 was used in excess to ensure that the polymer chains would be terminated with azides. For the largest polymer 3c, equimolar amounts of 1 and 2 were used during the polymerization. However, excess of 2 was added, along with more CuSO₄·5H₂O and ascorbic acid after the polymerization was complete, to ensure the fidelity of the end groups. These polymerizations yielded a series of high-molecular-mass polymers with polydispersity indices (PDIs) ranging from 1.71–1.90 (Table 1 and Fig. 2). These results are consistent with the expected step-growth polymerization mechanism, although lower PDIs have been observed for some polymers prepared by using A–A + B–B CuAAC polymerizations (30).

Table 1.

Gel permeation chromatography (GPC) of polymer thread 3(a–c) and polyrotaxane 5(a–c)·4nPF₆ (in parentheses)

Polymer sample	Nalkyne/Nazide	Retention volume, ml	Polydispersity index	Mn, kDa
3a (5a)	0.905	24.1 (25.0)	1.90	32
3b (5b)	0.975	23.3 (24.9)	1.78	53
3c (5c)	1.000	22.3 (23.3)	1.71	181

The number-average molecular masses (M_n) of 3(a–c) were determined by multiangle light scattering.

Fig. 2.

Gel permeation chromatographs of the polymer thread 3a and polyrotaxane 5a·4nPF₆. Although the molecular mass of 5a·4nPF₆ is nearly twice that of 3a, the retention volume of the polyrotaxane increases, a phenomenon attributed to its folded structure. Comparisons between polymer threads 3b and 3c and polyrotaxanes 5b·4nPF₆, 5c·4nPF₆, respectively, showed the same trend [see Table 1 and supporting information (SI) Fig. S1].

When 0.6 equivalents of CBPQT⁴⁺ relative to the total number of DNP units was added to solutions of the polymer threads 3a–c, the characteristic charge-transfer absorption band at 485 nm was observed by UV/Vis spectroscopy, a phenomenon that indicates the formation of the pseudopolyrotaxane [3⊂nCBPQT]·4nPF₆. The UV/Vis spectra of these solutions changed continually over the course of 24 h, after which time they ceased to change. These observations suggest that the threading process occurs slowly, an observation that was not surprising at all to us. Threading may occur only at the chain ends and previously threaded tetracationic cyclophanes presumably must migrate to the interior of the chains to free more unoccupied DNP units near the ends of the chains to bind additional rings. Finally, bulky 2,6-diisopropylphenyl stoppers were attached to the chain ends by using the CuAAC reaction between the pseudopolyrotaxanes [3⊂nCBPQT]·4nPF₆ and the stopper 4, mechanically locking the dynamic pseudopolyrotaxanes and converting them into the corresponding polyrotaxanes 5a–c·4nPF₆. The progress of the stoppering reaction was monitored by IR spectroscopy by observing the disappearance of the azide endgroup stretch at 2103 cm⁻¹ (see Fig. S2). When stoppering was complete, the polymers were precipitated into aqueous EDTA solutions to remove the Cu catalyst, and the majority of unthreaded CBPQT⁴⁺ was removed by filtering the solution through a DNP-functionalized PS-coDVB cross-linked resin. Finally, the polyrotaxanes were precipitated from DMF into CHCl₃ to remove unreacted monomers, low-molecular-mass oligomers and trace cyclic species. The coverage of CBPQT⁴⁺ rings into the polyrotaxanes was determined to be 90%, 74%, and 58% of CBPQT⁴⁺ per repeat unit for samples 5a–c·4nPF₆, respectively, by ¹H NMR spectroscopy, as a result of comparing the integration of the CBPQT⁴⁺ α-proton resonances with those of the collective glycol methylene resonances of the polymer backbone found between 3.0 and 5.0 ppm (see Fig. S4).

Results and Discussion

Polymer Characterization.

Gel permeation chromatography (GPC) analysis of the polyrotaxanes suggests that the polyrotaxanes 5a–c·4nPF₆ exhibit smaller hydrodynamic volumes than their direct synthetic precursors 3a–c. For example, the retention volume (Fig. 2) of 5a·4nPF₆ (25.0 ml) is larger than that of 3a (24.1 ml). Similar results were obtained (Table 1 and Fig. S1) for the retention volumes of the polymer threads 3b and 3c and the polyrotaxanes 5b·4nPF₆ and 5c·4nPF₆, respectively. These observations were particularly surprising because the addition of one CBPQT⁴⁺ ring per two DNP units would almost double the molecular mass of the polymer. This shift to higher retention volume is, however, consistent with the formation of a folded polymer conformation. This phenomenon is well known in the comparison of GPCs of globular dendrimers (31) compared with linear polymers of similar molecular mass and in intramolecularly cross-linked polymer chains (32).

Folding of Oligomeric Model Compounds.

A series of ¹H NMR spectroscopic studies on a model [2]rotaxane 6·4PF₆ (Fig. 3a) sheds some light on the folding behavior of the polyrotaxanes. The [2]rotaxane is a model compound for the polyrotaxane 5·4nPF₆ in which n = 1. It is also a triazole-containing analogue of the [2]pseudorotaxane [7⊂CBPQT]·4PF₆ (Fig. 3b) composed of a linear tri-DNP thread and a CBPQT⁴⁺ macrocycle, which has been shown (33) to fold into a similar π-stacked structure. This folding is quite apparent in the solid-state superstructure (Fig. 3c) of the complex, in which the π electron-donating CBPQT⁴⁺ encircles the middle DNP unit, and the other two DNP units stack on the outside of the tetracationic cyclophane, maximizing the donor–acceptor interactions by forming a five-layer, π-stacked donor–acceptor system.

Fig. 3.

Structural formula for the [2]rotaxane 6·4PF₆ (a), the [2]pseudorotaxane [7⊂CBPQT]·4PF₆ (b), and the x-ray crystal structure of [7⊂CBPQT]⁴⁺ (c).

The ¹H NMR spectra of the polyrotaxane 5c·4nPF₆ and the model [2]rotaxane 6·4PF₆ are quite similar (Table 2 and Table S1) to those reported (33) for the [2]pseudorotaxane [7⊂CBPQT]·4PF₆, suggesting that the CuAAC-derived compounds adopt a very similar folded secondary structure. In the spectrum (Fig. 4a) of 6·4PF₆ recorded at 245 K, the α-CH, β-CH, and N-methylene protons signals of the CBPQT⁴⁺ ring, are shifted upfield (Δδ = 0.33–1.54 ppm) relative to free CBPQT⁴⁺, as expected when the cyclophane is encircling a DNP unit. Each of the signals is split into two resonances, and this observation, combined with the overall symmetry of the spectrum, is suggestive of a structure in which the local symmetry is C_2h, with the CBPQT⁴⁺ located on the middle DNP unit. The signals of the central DNP proton resonances (H-2/6, H-3/7 and H-4/8) were shifted upfield (Δδ = 1.12–5.97 ppm) to a degree diagnostic for CBPQT⁴⁺ binding. Furthermore, the alongside DNP protons (H-2′ − H-4′, H-6′ − H-8′) exhibit six distinct signals, each shifted upfield to a smaller extent (Δδ = 0.11–0.64 ppm) than those of the central unit. The assignment of all of the above protons were also confirmed (Fig. S3) by COSY ¹H NMR spectroscopy at 245 K. The shifts of these resonances to higher field are attributed to the secondary donor–acceptor interactions that cause the alongside DNP units to stack on the outside of the CBPQT⁴⁺ rings. Thus, the direct comparison of the symmetry and magnitude of the chemical shifts in the [2]rotaxane 6·4PF₆ and the [2]pseudorotaxane [7⊂CBPQT]·4PF₆ suggests that they fold in similar ways.

Table 2.

Selected induced 500-MHz ¹H NMR chemical shift changes for the [2]pseudorotaxane [7⊂CBPQT]·4PF₆, the [2]rotaxane 6·4PF₆, and polyrotaxane 5c·4nPF₆ in CD₃CN solution

Proton	[2]Pseudorotaxane [7⊂CBQT]·4PF6			[2]Rotaxane 6·4PF6			Polyrotaxane 5c·4nPF6
	δ uc	δ c (243 K)	Δ δ	δ uc	δ c (245 K)	Δ δ	δ c (245 K)
CBPQT4+
α -CH	8.86	8.68	−0.18	8.86	8.50	−0.36	8.45
		8.12	−0.74		8.25	−0.61	8.19
β -CH	8.16	6.63	−1.53	8.16	6.79	−1.37	6.73
		6.57	−1.59		6.62	−1.54	6.55
DNP
Alongside H-4′	7.77	7.32	−0.45	7.76	7.12	−0.64	7.23
Alongside H-8′	7.79	7.33	−0.46	7.79	7.38	−0.41
Alongside H-7′	7.34	7.16	−0.18	7.38	7.27	−0.11	7.10
Alongside H-3′	7.32	7.07	−0.25	7.35	7.01	−0.34
Alongside H-6′	6.88	6.69	−0.17	6.89	6.74	−0.15	6.54
Alongside H-2′	6.82	6.40	−0.42	6.88	6.56	−0.32

δ _uc is the chemical shift in the spectrum of the protons of each component in the unencircled state at 245 K. δ _c is the chemical shift of the protons observed in the encircle systems. Δ δ = δ _c = δ _uc. The data for [2]pseudorotaxane [7⊂CBPQT]·4PF₆ are cited from ref. 33. A complete list of chemical shifts is shown in Table S1.

Fig. 4.

VT ¹H NMR spectroscopic studies. (a) The [2]rotaxane 6·4PF₆. Separate groups of α, β, and C₆H₄ protons are distinguishable at room temperature and below. Further sharpening of the peaks and splitting caused by coupling are observed when the sample is cooled down to 245 K. (b) The polyrotaxane 5·4nPF₆. The spectrum of polyrotaxane has peaks with chemical shifts identical to those reported for the [2]rotaxane 6·4PF₆. Coalescence of peaks is also observed while warming up the sample.

The polyrotaxane 5·4nPF₆ displayed very similar NMR spectra (Fig. 4b) when compared with those of [2]rotaxane 6·4PF₆(Fig. 4a). Protons on the CBPQT⁴⁺ ring, and on the included DNP unit, exhibit almost identical chemical shifts and splitting patterns to those observed for model compounds. In the polymer constitution, there is little difference in chemical environment between H-2′ and H-6′, H-3′ and H-5′, and H-4′ and H-8′, respectively, alongside the DNP units, and so these resonances are not split into individually assignable peaks as in the models 6·4PF₆ and [7⊂CBPQT]·4PF₆. With the exception of this difference, the ¹H NMR spectrum of the polyrotaxane recorded at 245 K is quite similar to that of the [2]rotaxane 6·4PF₆, exhibiting the same splittings for the CBPQT⁴⁺ α-CH, β-CH, aromatic C₆H₄, and methylene proton resonances, as well as changes in the chemical shifts of the included in and alongside the DNP residues that are indicative of the folded conformation.

VT NMR spectroscopy was also used to probe the relative stabilities of the folded nature of the model and the polymer (34). At higher temperatures, the unfolding and subsequent refolding process becomes rapid on the NMR time-scale, resulting in the coalescence of the CBPQT⁴⁺ α-CH, β-CH, aromatic C₆H₄ and methylene proton resonances. The activation barrier for this process is therefore a measure of the relative stability of the folded conformation (35). Observation of the coalescence of signals arising from the CBPQT⁴⁺ α-CH, β-CH, aromatic C₆H₄ protons in the spectra of 6·4PF₆ (Table S2) afforded an average value for the activation barrier of 12.0 kcal·mol⁻¹, indicating that the folded structure is somewhat less kinetically stable than in [7⊂CBPQT]·4PF₆, whose ΔG^‡ value is 14.7 kcal mol⁻¹. Interestingly, in the spectra of the polyrotaxanes 5·4nPF₆,the signals for the α-CH protons at 8.2–8.7 ppm and the β-CH protons at 6.6–6.9 ppm, respectively, did not coalesce, even when the solution was heated to 335 K, i.e., just below the lower critical solution temperature of the polymer. This observation points to a cooperative effect in the thermal stability of the folded structure of the polymeric system. In the semicontinuous π–π stacked folded structure hypothesized for the polyrotaxane, the process by which these resonances interconvert most likely requires the simultaneous unfolding of two or more folded units, a much more energy-intensive process than the unfolding process in a [2]rotaxane.

Atomic Force Microscopy.

The application of atomic force microscopy (AFM) (36) to the observation of individual polymer molecules enables real-space imaging of structural changes resulting from polymer folding (37–39). Measurements of both the polymer thread and polyrotaxane have revealed distinct differences in the appearance and overall dimensionality through cross-sectional analyses. Fig. 5 provides representative AFM images of both 3c and 5c·4PF₆ after deposition onto highly ordered pyrolytic graphite (HOPG). Optimization of the deposition conditions was carried out to minimize intermolecular interactions and produced samples with mostly isolated structural units. The determination of these observed structures as single molecules was carried out through the use of cross-sectional heights, as shown in the respective insets illustrated at the bottom of Fig. 5 a and b.

Fig. 5.

Representative AFM height images of polymer thread (a), polyrotaxane (b), and cross-sectional linescans (Insets) on HOPG. Scan parameters: size,1 μm²; scan rate, 0.5 Hz.

Variations in the observed feature length of the polyrotaxane, compared with the polymer thread, provide further evidence for the folding phenomenon. Specifically, a decrease in length and corresponding increase in both height and width for the polyrotaxane were observed (Table 3). The decrease in length and increase in width can be attributed to the longitudinal compression of the polymer chain and transverse structural broadening expected in the folding process (40). An increase in the measured height is attributed to the addition of CBPQT⁴⁺ rings along the backbone of the polyrotaxane. Statistical analyses of height and width values from multiple samples, images, and tips determined the two systems to be statistically different to a 99.995% confidence level by using the two-sample t test. The single atomic step height of HOPG was used as an additional calibration in height measurements. Because of large standard deviations, compiled data were plotted as a histogram and fitted by using nonlinear least-squares analysis for increased accuracy. This type of presentation of the data was especially useful in the analysis of the molecular lengths, which exhibited a bimodal structure (Fig. 6 and Table S3) in the plotted histogram and was not readily represented by a simple mean value.

Table 3.

Observed relative dimension changes resulting from the folding process

	Δ %L	Δ %W	Δ %H
3c → 5c·4nPF6	(−) 15.64	92.06	43.18

Negative values, denoted by (−) represent a decrease in measured molecular dimensions.

Fig. 6.

Histogram and fitting results of measured molecular lengths from cross-sectional analyses demonstrating a decrease in structural length upon folding.

The observed variation in molecular dimensions compared with expected values from structural modeling and crystallography, can be attributed to image convolution resulting from both the finite radius of the probe tip and loading force applied to the sample to maintain stable, controlled feedback (41).

Conclusions

In summary, we synthesized DNP–CBPQT⁴⁺-based polyrotaxanes, using the CuAAC, first for a step-growth polymerization of two different DNP units and then for the final stoppering step in the threading-followed-by-stoppering process that serves to constrain the CBPQT⁴⁺ rings in the dumbbell component of the polyrotaxane. The polyrotaxanes adopt an alternating stacking pattern that is induced by secondary noncovalent bonding interactions between the rings and the alongside DNP units. GPC studies show that the folding causes a significant decrease in the polymer dimensions despite a near twofold molecular mass increase—a unique phenomenon that sets the polyrotaxane apart from traditional polymeric materials. The folding behavior was also confirmed by ¹H NMR spectroscopy by comparing the polyrotaxanes with a simple folded [2]rotaxane model compound. Moreover, analysis of individual polymer chains dispersed on HOPG by AFM indicates that the polyrotaxanes exhibit increased height and width but decreased lengths compared with the precursor threads, a result that is consistent with the other characterization methods. This research provides a strategy for the controllable fabrication of materials with elaborate structures at the nanometer-scale level. Directions for the research to take in the future could include the development of switchable polymeric systems incorporating multiple recognition units, which might find uses in nanoelectromechanical systems (NEMS) and molecular electronic devices (MEDs).

Materials and Methods

Chemicals were purchased from commercial suppliers and were used as received. BHEEN (27), bis(2-(2-azidoethoxy)ethoxy)-naphthalene (13), CBPQT·4PF₆ (27) were prepared according to literature procedures. Gel permeation chromatography (GPC) was carried out on 15-μm columns set up in series with two linear mixed beds, followed by a 100-Å column. A 0.5 M solution of NH₄PF₆ in DMF was used as the mobile phase at 0.500 ml/min flow from an Agilent 1100 isocratic pump. The detector system consisted of a miniDawn three angle, light-scattering system, followed downstream by an Optilab Rex differential refractometer from Wyatt Technologies. Number-average molecular mass (M_n), weight-average molecular mass (M_w) and polydispersities (PDI = M_w/M_n) were determined relative to linear polystyrene (GPC_PSt). ¹H and ¹³C NMR spectra were recorded on a Bruker DRX500 (500 MHz) or AV600 (600 MHz) spectrometer, with the residual solvent resonance as the internal standard. Infrared (IR) spectra were measured on a Perkin–Elmer 100 FT/IR spectrometer by using AgCl plates as the sample window. Electrospray ionization (ESI) mass spectra were measured on an Ionspec Fourier Transfer HiRes ESI Mass Spectrometer by using MeCN as the mobile phase. Electron ionization (EI) mass spectra were measured on a Shimadzu QP2010S GC-EI Mass Spectrometer. Fast atom bombardment mass spectra (FAB) were obtained on a JEOL JMS-600H high-resolution mass spectrometer equipped with a FAB probe.

General Polymerization Procedure.

Polymer Thread 3: Monomers 1 (290 mg, 702 mmol) and 2 (3a, 1:2 = 0.905:1; 3b, 1:2 = 0.975:1; 3c, 1:2 = 1:1), ascorbic acid (27.3 mg, 0.155 mmol) and CuSO₄·5H₂O (19.4 mg, 0.0776 mmol) were dissolved in DMF (4 ml). The mixture was stirred under argon for 24 h. In the case of 3c, another 3 mg of 2 was added to the solution along with 0.1 mg of ascorbic acid and 0.1 mg of CuSO₄·5H₂O, and the reaction mixture was stirred for an additional 3 h. The mixture was then added dropwise to an aqueous EDTA solution (0.1 M, 20 ml), forming a yellow precipitate. The precipitate was collected by centrifugation before being washed with H₂O (20 ml times two) and dried under high vacuum. The resulting solid was dissolved in CH₂Cl₂ (2 ml) and precipitated into Me₂CO (20 ml) to give the product. The solid was again recovered by centrifugation and finally dried under high vacuum. ¹H NMR (500 MHz, CDCl₃): δ = 8.08–8.04 (m, 2H, Triazole-H), 7.84–7.71 (m, 4H, Ar-H^2/6), 7.42–7.30 (m, 4H, Ar-H^3/7), 7.05–6.84 (m, 4H, Ar-H^4/8), 4.61–3.60 (m, 32H, CH₂O), 3.50–3.42 (m, 4H, CH₂-Triazole).

Synthesis of Polyrotaxane 5·4nPF₆.

A solution of CBPQT·4PF₆ (48.1 mg, 0.0437 mmol) and polymer thread 3 (29.1 mg, N_DNP = 0.0728 mmol) in DMF (2 ml) was stirred at room temperature for 24 h. A purple solution indicative of pseudopolyrotaxane [3⊂nCBPQT]·4nPF₆ formation resulted. Stock solutions containing stopper 4 (1.6 mg, 0.00725 mmol), ascorbic acid (0.1 mg, 0.000726 mmol), and CuSO₄·5H₂O (0.1 mg, 0.000363 mmol), were then added to the solution. The reaction mixture was stirred under Ar at room temperature overnight. Resin-supported DNP S6 (see SI Text) was added to the solution for 10 min before being removed by filtration. The solution was then added to an aqueous EDTA solution (0.1 M, 5 ml), forming a purple precipitate. The precipitate was recovered by centrifugation, washed with H₂O (5 ml times two) and dried under high vacuum for 2 h. The solid was dissolved in a minimum amount of DMF and precipitated when added to CHCl₃ (5 ml). The resulting polyrotaxane was again recovered by centrifugation and dried under vacuum. For further experimental details, see SI Text and Schemes S1-S5.